Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semantic Evolvement Enhanced Graph Autoencoder for Rumor Detection

Xiang Tao, Liang Wang, Qiang Liu, Shu Wu, Liang Wang

TL;DR

The paper addresses misinformation spread by integrating semantic evolvement into rumor detection. It introduces GARD, a semantic evolvement enhanced Graph Autoencoder that learns local semantic changes along propagation paths and global evolvement via masked feature reconstruction, augmented by a uniformity regularizer. The model optimizes a joint objective combining supervised loss with self-supervised reconstruction losses, achieving superior results on Twitter15, Twitter16, and PHEME, especially for early detection. This approach improves robustness and generalization while reducing reliance on complex data augmentation, offering practical benefits for timely and reliable rumor mitigation.

Abstract

Due to the rapid spread of rumors on social media, rumor detection has become an extremely important challenge. Recently, numerous rumor detection models which utilize textual information and the propagation structure of events have been proposed. However, these methods overlook the importance of semantic evolvement information of event in propagation process, which is often challenging to be truly learned in supervised training paradigms and traditional rumor detection methods. To address this issue, we propose a novel semantic evolvement enhanced Graph Autoencoder for Rumor Detection (GARD) model in this paper. The model learns semantic evolvement information of events by capturing local semantic changes and global semantic evolvement information through specific graph autoencoder and reconstruction strategies. By combining semantic evolvement information and propagation structure information, the model achieves a comprehensive understanding of event propagation and perform accurate and robust detection, while also detecting rumors earlier by capturing semantic evolvement information in the early stages. Moreover, in order to enhance the model's ability to learn the distinct patterns of rumors and non-rumors, we introduce a uniformity regularizer to further improve the model's performance. Experimental results on three public benchmark datasets confirm the superiority of our GARD method over the state-of-the-art approaches in both overall performance and early rumor detection.

Semantic Evolvement Enhanced Graph Autoencoder for Rumor Detection

TL;DR

The paper addresses misinformation spread by integrating semantic evolvement into rumor detection. It introduces GARD, a semantic evolvement enhanced Graph Autoencoder that learns local semantic changes along propagation paths and global evolvement via masked feature reconstruction, augmented by a uniformity regularizer. The model optimizes a joint objective combining supervised loss with self-supervised reconstruction losses, achieving superior results on Twitter15, Twitter16, and PHEME, especially for early detection. This approach improves robustness and generalization while reducing reliance on complex data augmentation, offering practical benefits for timely and reliable rumor mitigation.

Abstract

Due to the rapid spread of rumors on social media, rumor detection has become an extremely important challenge. Recently, numerous rumor detection models which utilize textual information and the propagation structure of events have been proposed. However, these methods overlook the importance of semantic evolvement information of event in propagation process, which is often challenging to be truly learned in supervised training paradigms and traditional rumor detection methods. To address this issue, we propose a novel semantic evolvement enhanced Graph Autoencoder for Rumor Detection (GARD) model in this paper. The model learns semantic evolvement information of events by capturing local semantic changes and global semantic evolvement information through specific graph autoencoder and reconstruction strategies. By combining semantic evolvement information and propagation structure information, the model achieves a comprehensive understanding of event propagation and perform accurate and robust detection, while also detecting rumors earlier by capturing semantic evolvement information in the early stages. Moreover, in order to enhance the model's ability to learn the distinct patterns of rumors and non-rumors, we introduce a uniformity regularizer to further improve the model's performance. Experimental results on three public benchmark datasets confirm the superiority of our GARD method over the state-of-the-art approaches in both overall performance and early rumor detection.
Paper Structure (22 sections, 18 equations, 4 figures, 4 tables)

This paper contains 22 sections, 18 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Rumor Detection
  4. Graph Autoencoders
  5. Problem Definition
  6. The proposed GARD Model
  7. Local Semantic Evolvement Learning
  8. Global Semantic Evolvement Learning
  9. Representation of Propagation Graph
  10. Uniformity Regularizer
  11. Experiments
  12. Evaluation Setups
  13. Datasets
  14. Baselines
  15. Experimental Settings
  16. ...and 7 more sections

Figures (4)

  • Figure 1: (a) the top-down semantic evolvement graph, as comments increase, semantic begin to evolve, (b) the reverse bottom-top semantic evolvement graph, where the edges between nodes indicate the direction of features reconstruction.
  • Figure 2: The overall framework of our proposed GARD model. (1) The learning of local semantic changes is achieved by reconstructing node features in both the top-down and bottom-up directions of parent-child node pairs. (2) The learning of global semantic evolvement is achieved by conducting features random mask reconstruction on undirected propagation graph. (3) We introduce a uniformity regularizer to enhance the model's ability to learn the distinctive patterns of events.
  • Figure 3: Sensitivity analysis of hyperparameters $\alpha_1$ and $\alpha_2$, which represent the weight of reconstructed loss and uniformity loss.
  • Figure 4: Results of rumor early detection task on three datasets